U.S. Energy Secretary Rick Perry speaks at an energy summit in Salt Lake City in May. On Monday, Perry attended a presentation on artificial intelligence at Lawrence Livermore National Laboratory in Livermore. (AP Photo/Rick Bowmer) Artificial intelligence, in the form of rapid collection and analysis of massive sets of biologic data, already has revolutionized medicine – by some estimates, the world’s fastest machines have processed more data in the past two years than in all of human history. But the technology needs to be much more powerful if it’s going to have real-world consequences for people with everything from traumatic brain injuries to cancer, kidney disease and Alzheimer’s.
That was the message Monday at a meeting of the country’s top scientists in brain and computational research at Lawrence Livermore National Laboratory, after which U.S. Energy Secretary Rick Perry signed an agreement with a major philanthropic foundation to promote formal partnerships between the Department of Energy and public and private institutions around the country.
Exactly what types of ventures the memorandum of understanding between the DOE and the Weill Family Foundation, based in New York, will support are not yet known. But the agreement could bolster a relatively new partnership between UCSF and scientists from the national labs in Livermore and Berkeley that is focused on processing enormous amounts of data from people with traumatic brain injuries, in hopes of developing better diagnostic equipment and even treatments.
“We have all this complex data, and the fact is that our current computing infrastructure doesn’t support us making the data as actionable as we would like,” said Dr. Geoffrey Manley, a neurosurgeon who leads traumatic brain injury research at UCSF. “Where this partnership (with the national labs) provides value, is these folks apply machine learning and AI and brute force computation that we really haven’t had access to.”
Specifically, the UCSF neuroscientists were able to work with computer scientists and engineers – using the Sierra supercomputer in Livermore, the second-fastest computing system in the world – to transform analysis of data collected from hundreds of patients as part of an ongoing traumatic brain injury study, which Manley co-heads. One particular computation that used to take the UCSF scientists nine hours to complete was done in two minutes and 37 seconds at the Livermore lab, Manley said.
Perry said at the start of Monday’s meeting that partnerships like the one between UCSF and the laboratories – plus further collaborations with private industry – are going to be critical to “change the world” with artificial intelligence. “We can revolutionize treatments for everything from cancer to brain injury. The potential here is staggering,” Perry said. “I’m not sure any of us really has a grasp of just how broad and deep the impact can be.”
Artificial intelligence, as it applies to medicine, is essentially about taking the brain’s natural ability for studying data and identifying trends and powering it way up. Medical scientists have had access to enormous datasets for decades, but only in the past 10 or 20 years have computers existed that were powerful enough to analyze that data at a speed that was useful. Related Stories
The amount of biologic data that is available for research is unimaginable, scientists at Monday’s meeting said. Every day, health care providers, academic institutions, and even mobile apps are amassing information that could transform medicine by identifying nuanced trends in population-level health. Human scientists alone could never process that amount of data – just coming up with the right questions to ask has proved daunting.
“Biomedical is the most complex dataset on the planet,” said Keith Yamamoto, vice chancellor for science policy at strategy at UCSF. “There is a lot of data. And it’s growing.”
Creating computer systems capable of processing the increasing data load is an ongoing scientific and engineering problem – which is partly why public-private partnerships are going to be invaluable moving forward, industry leaders said Monday. The private-sector scientists can build systems that might be capable of running data sets in just minutes that would take academic researchers weeks, or longer. By providing them with speedier, more powerful equipment, the engineers also give medical researchers more freedom – in the form of time – to test new theories and explore.
“If you take work that currently takes a month and make it a few minutes, you allow your scientist to test thousands of new ideas,” said Andrew Feldman, co-founder of Cerebras Systems in Los Altos. “That’s the glory of being an infrastructure builder – we don’t have new ideas, we get great pleasure from building a platform onto which your ideas can take shape.”
Perry and others at the Livermore meeting said another motivator for forging public-private partnerships is competition – specifically from China, which is neck-and-neck with the United States in AI development, scientists said.
China has somewhat of an advantage because its government, academic and private sectors already have close connections – that’s the upside of not being a democracy, said Michael Pecht, director of the Center for Advanced Life Cycle Engineering at the University of Maryland.
“We’re pretty even now, but China is trending stronger,” Pecht said.
Erin Allday is a San Francisco Chronicle staff writer. Email: [email protected]